Temporal Methods Seminar - ASTM International · Presentation of the problem and logistics ... Aromatic Profile of Prototype E18 Bars ... reference is sucrose in water. Temporal Methods

Temporal Methods Seminar April 2013

Temporal Methods Seminar

1:00 to 5:00 PM Wednesday April 17th 2013

Seminar Leaders: Suzanne Pecore General Mills Inc Chris Findlay Compusense Inc Sarah Kirkmeyer Givaudan Inc.



Time is an important factor in the sensory experience. The onset,

duration and sequence of sensations evoked by human sensory

interaction with products are key properties in product performance

and differentiation. To be able to understand these temporal

properties we must first be able to measure them. At this time, ASTM

has only one document relating to temporal measures, ASTM E1909-

11 Standard Guide for Time-Intensity Evaluation of Sensory

Attributes.

I max

T max Dur

AUC

Inc AUC Dec AUC

Dec Angle Inc Angle

STANDARDIZED PARAMETERS FROM T-I CURVES

Tmax – time at maximum intensity

Imax – maximum intensity

DUR – duration

AUC – area under curve

Inc. Angle – increase angle

Inc. Area – increase area

Dec. Angle – decrease angle

Dec. Area – decrease area

Time Intensity Parameters

Analyzing Curve Parameters



A number of newer methods have been developed and are being used

by sensory researchers around the world. In the interest of learning

more about these methods and considering their inclusion in the

ASTM publications, this seminar will focus on three of those methods

and provide participants with a “Hands-on” experience with real

products and real data collection.

•

–

–

–

•

–

–

–

–



The three methods are:

1. Progressive Profiling

Chris Findlay and Miranda Robb

2. TDS (Temporal Dominance of Sensations)

Sarah Kirkmeyer and John Castura

3. TOS (Temporal Order of Sensations)

Suzanne Pecore and Peter Love

Everyone will have an opportunity to rotate through and

experience all three methods.


The final part of the seminar will be devoted to analysis, interpretation

and evaluation of the strengths, weaknesses and challenges of these

temporal methods. Participants will be able to determine the ease of use

by panelists, the clarity of results, the skills required for the analyst and

finally the potential value to their business.

E18 will be able to consider developing a new document or documents on

Temporal Methods


Schedule: 1:00 Introduction

Presentation of the problem and logistics

1:30 Hands-on #1

2:10 Hands-on #2

2:50 Hands-on #3

3:30 Break for Participants

Data Analysis by group leaders

4:00 Presentation of results

4:30 Discussion of the methods and observations


Application of Temporal Methods Problem Statement

Project Objective

• The E18 Snack Bar is being reformulated to improve its health credentials.

• Goal is to match the sensory characteristics of the current E18 Bar as closely as possible.


Descriptive Analysis shows options with similar aromatics to current bar.

0

1

2

3

4

5

6

Mea

n r

atin

g (1

5 p

t in

ten

sity

sca

le)

Aromatic Profile of Prototype E18 Bars

Control Option 1 Option 2 Option 3 Option 4

ab ababb

a

Bars sharing the same letter do not significantly differ in intensity at the 95% confidence level.

Descriptive Analysis also shows potential issues with the temporal taste profile.

0

1

2

3

4

5

6

Sweet 60 Sec Sweet Salt Sour 60 Sec Sour Bitter 60 Sec Bitter

Mea

n R

atin

g (1

5 p

t in

ten

sity

sca

le)

Tastes/Aftertastes of Prototype E18 Bars

Control Option 1 Option 2 Option 3 Option 4

ab

abab

b

a

Bars sharing the same letter do not significantly differ in intensity at the 95% confidence level.

Option 2, for example, is less sweet in-mouth, but with more lingering sweet taste.

Temporal Study

• Objective is to understand if the temporal profile has been affected by the reformulation of the E18 bar, and

• To identify a formulation option that more closely matches the Control bar.


Attributes To Be Evaluated By The Temporal Methods

Attributes Definition

Grain Aromatics associated with grains such as oat, wheat, etc. An overall grain aromatic, inclusive of raw, cooked and toasted aspects.

Caramelized Aromatic associated with caramelized sucrose or browned sugars

Dried Fruit A brown sweet aromatic associated with dried fruits such as raisins and prunes.

Nutty Aromatics associated with nut meats such as almonds. An overall nutty aromatic, inclusive of raw and roasted aspects.

Sweet Taste on the tongue stimulated by sugars and potent sweeteners; reference is sucrose in water.


Evaluation Protocol

• Bite through the bar with incisors, about ½ inch into bar.

• Try to sample a representative section of bar, i.e., avoid large particulates. (Can break-up bar to reach representative sections.)

• Chew bar with molar teeth.

• Start of timing should coincide with first chew. (Details provided at each station.)

• Expectorating is encouraged!


Logistics • Three samples: 227, 648, 950

• Evaluated as two pairs:

Group 1 (Blue circle) : 227 vs. 950

Group 2 (Red triangle): 227 vs. 648

• You are assigned to Group 1 or Group 2

• Each Group will evaluate the same pair at each of three stations:

Station 1 (A) Room 307: Progressive Profiling

Station 2 (B) Room 308: TDS

Station 3 (C) Room 309: TOS

• Observations collected at each station




Progressive Profiling

Chris Findlay Miranda Robb Compusense Inc


Attributes to be evaluated by the temporal methods









Progressive Profile Ballot




Observations


• Thoughts / Insights as you experience this approach?

• Build on comments from prior groups (see flip charts)

LET’S COLLECT SOME DATA



Temporal Order of Sensations

Session Leaders: Suzanne Pecore General Mills Inc Peter Love Compusense Inc

What is TOS?

• TOS is a technique to measure the order that key attributes appear over the eating experience, i.e., over several spoonfuls and into the aftertaste


Evaluation Protocol

• Bite through the bar with incisors, about ½ inch into bar.

• Try to sample a representative section of bar, i.e., avoid large particulates. (Can break-up bar to reach representative sections.)

• Chew bar with molar teeth.

• Start of timing should coincide with first chew.

• Expectorating is encouraged!


Specifics of E18 Bar Evaluation • You will see one of two pairs:

Group 1: 227 vs. 648

Group 2: 227 vs. 950

• Focus is to identify the appearance, in order, of the first three attributes from a possible six attributes.

• Repeated THREE TIMES

• 4th bite is chewed 10-12 times, then most intense aftertaste is selected 10 seconds after expectoration.


Attributes To Be Evaluated







Other Describe


Questionnaire for TOS

Observations


• Thoughts / Insights as you experience this approach?

• Build on comments from prior groups (see flip charts)

Introduction to TDS Sarah V. Kirkmeyer, Givaudan

John Castura, Compusense

ASTM - April 2013


April 2013

TDS Definition

The Temporal Dominance of Sensations is a sensory profiling technique

The aim of the TDS methodology is to give an overall temporal picture of a product by

allowing the panelist to score several descriptors simultaneously (multi-attributes)

It detects sequence of dominance of different attributes

(intensity of the dominant attribute at the time ti is recorded)

At Givaudan, TDS methodology is based on Quantitative Flavor Profiling (QFP) for

descriptive part and TI for temporal information


April 2013

What kind of information do we obtain?

The sequence of dominance in the profile

The time and the duration when each attribute starts to be dominant (until it changes

to another one)

The intensity of each dominant perception *note requires significant additional training

Is there the dominance of an attribute disappearing and reappearing later? (“Nutty” in

the example)

Pineau, Pessina, Cordelle, Imbert,

Rogeaux, Schlich,

(7th Sensometrics Meeting, Davis,

CA, 2004)

General Training sessions

Learning instruments/descriptive attributes

Recognize instruments/descriptive attributes quickly at the first note/perception

Learning sequencing

Check that sequencing and time dependent is understood

Learn sequencing taking into account the intensity

Defining dominance

Sequence then scoring

Dominance, NOT intensity

Dominance scale is a proportion


April 2013

The process with food products

Defining language used

Defining in mouth and after swallowing profile, protocol of tasting

Defining temporality flavor beginning/middle/end

Checking time of shift

Checking score


April 2013

Interpreting Results

6

Remember Dominance, NOT intensity

Dominance scale is a proportion

Significance line keeps the focus on the meaningful output

Do not create chances for obsessing over noise

Highlight the main points/conclusions

Useful to compare two TDS curves side by side when comparing samples or products


April 2013

Attributes To Be Evaluated


Grain Aromatics associated with grains such as oat, wheat,

etc. An overall grain aromatic, inclusive of raw, cooked

and toasted aspects.

Caramelized Aromatic associated with caramelized sucrose or

browned sugars

Dried Fruit A brown sweet aromatic associated with dried fruits

such as raisins and prunes.

Nutty Aromatics associated with nut meats such as almonds.

An overall nutty aromatic, inclusive of raw and roasted

aspects.

Sweet Taste on the tongue stimulated by sugars and potent

sweeteners; reference is sucrose in water.


April 2013

What do we ask the panel?

Observations


April 2013

Thoughts / Insights as you experience this approach?

Build on comments from prior groups (see flip charts)


Temporal Methods Seminar Conclusion 4:00 to 5:00 PM

Wednesday April 17th 2013

Break out Groups: Progressive Profile Chris Findlay Compusense TDS Sarah Kirkmeyer Givaudan TOS Suzanne Pecore General Mills


The intent of the final session in the seminar:

Analysis, interpretation and evaluation of the strengths, weaknesses and

challenges of these temporal methods, the ease of use by panelists, the

clarity of results, the skills required for the analyst and finally the

potential value to any business.

The outcome for ASTM:

E18 will be able to commence developing a new document or documents

on Temporal Methods.


Data Analysis

• Each data point is made up of the 5 well-defined attributes whose intensity has been scored on an unstructured line-scale anchored at 0 and 100

• Timing is achieved by specific pauses between data points. A well-trained and calibrated panel will be recording data at very similar times. Our seminar panelists have been disrupted by connectivity problems, lack of practice and distraction.

• The time points were planned to be as follows – Initial point about 10 seconds

– Point 2 about 30 seconds

– Point 3 20 seconds after 2nd Bite

– Point 4 20 seconds after 3rd Bite

– Aftertaste 30 seconds after swallowing or spitting the 3rd Bite

• Total time elapsed, about 3 minutes

• Exact time stamps may be collected to achieve individual timing of panelists for training and analysis. Analysis has yet to be developed for this data.



Test Name: Progressive Profiling 227 Combined from Both Groups

# of Evaluations: 46 23 Per Group

Summary Results

Multiple Comparison Test Used: Tukey's HSD 0.05

Attribute Title time p value HSD value 950 227 648

Grain Flavor 1 0.06 9.96 50.78 a 41.15 a 41.91 a

Caramelized Flavor 1 0.07 10.51 36.87 a 33.46 ab 25.43 b

Dried Fruit Flavor 1 0.11 12.36 47.7 a 37.67 a 36.78 a

Nutty 1 0.1 11.23 52.39 a 42.13 a 45.26 a

Sweet 1 0.05 9.42 56.91 a 47.13 b 49.26 ab

Grain Flavor 2 0.45 11.97 47.65 a 44.63 a 40.43 a

Caramelized Flavor 2 0.17 8.45 32.61 a 32.54 a 26.22 a

Dried Fruit Flavor 2 0.39 14.49 40.65 a 38.11 a 31.52 a

Nutty 2 0.38 12.28 48.57 a 41.43 a 43.83 a

Sweet 2 0.1 9.02 49 a 43.72 ab 39.65 b

Grain Flavor 3 0.41 8.12 44.78 a 45.54 a 41.09 a

Caramelized Flavor 3 0 7.62 43.96 a 27.2 b 28.78 b

Dried Fruit Flavor 3 0.01 10.5 47.83 a 37.3 b 30.91 b

Nutty 3 0.2 8.3 45.57 a 44.43 a 39 a

Sweet 3 0 6.34 57.04 a 47.48 b 36.57 c

Grain Flavor 4 0.02 8.14 50.39 a 43.04 ab 39.35 b


Dried Fruit Flavor 4 0.04 12.32 47.65 a 35.09 b 35.65 ab

Nutty 4 0.02 9.71 44.22 a 45.09 a 33.83 b

Sweet 4 0.01 7.15 50.3 a 43.39 ab 40.04 b

Grain Flavor 5 0.57 8.21 39.91 a 42.04 a 38.57 a


Dried Fruit Flavor 5 0.02 11.27 33.43 a 20.7 b 20.91 b

Nutty 5 0.42 11.53 35.91 a 35.46 a 29.7 a

Sweet 5 0.01 10.58 41.52 a 36.26 ab 25.96 b

Summary Analysis of Variance Produced immediately after data collection


RED Only evaluated 648 and 227

Attribute Title 648 227 p

1 Grain Flavor 41.91 36.3 0.13

1 Caramelized Flavor 25.43 28.52 0.50

1 Dried Fruit Flavor 36.78 38.04 0.83

1 Nutty 45.26 42.26 0.58

1 Sweet 49.26 49.48 0.96

2 Grain Flavor 40.43 40.26 0.98



2 Nutty 43.83 41.74 0.76

2 Sweet 39.65 39.91 0.94

3 Grain Flavor 41.09 40.96 0.97



3 Nutty 39 42.7 0.35

3 Sweet 36.57 40.96 0.21

4 Grain Flavor 39.35 40.87 0.71



4 Nutty 33.83 44.65 0.07

4 Sweet 40.04 43.61 0.27

5 Grain Flavor 38.57 39.04 0.90



5 Nutty 29.7 33.22 0.58

5 Sweet 25.96 32.22 0.28

BLUE Only evaluated 950 and 227

Attribute Title 950 227 p

1 Grain Flavor 50.78 41.15 0.43



1 Nutty 52.39 42.13 0.07

1 Sweet 56.91 47.13 0.02

2 Grain Flavor 47.65 44.63 0.82



2 Nutty 48.57 41.43 0.13

2 Sweet 49 43.72 0.79

3 Grain Flavor 44.78 45.54 0.20



3 Nutty 45.57 44.43 0.89

3 Sweet 57.04 47.48 0.54

4 Grain Flavor 50.39 43.04 0.24



4 Nutty 44.22 45.09 0.72

4 Sweet 50.3 43.39 0.08

5 Grain Flavor 39.91 42.04 0.23



5 Nutty 35.91 35.46 0.71

5 Sweet 41.52 36.26 0.82

The Blue Group (950 vs 227) and the Red Group (648 vs 227)


BLUE

BLUE

RED

RED


BLUE

BLUE RED

RED


BLUE RED


Two Controls (227) plotted together.

Preliminary Conclusions


• Both options, 950 and 648 , show significant differences from sample 227.

• Sample 227 in both Red and Blue groups show great similarity in all attributes at all time points.

• Sample 648 appears to be the closer option given that the only significantly different attribute was Nutty at time point 4.


TDS Conclusions

Sarah Kirkmeyer Givaudan Flavors John Castura Compusense

Objectives of TDS Evaluation

• To understand if the temporal profile has been affected by the reformulation of the E18 bar, and

• To identify a formulation option that more closely matches the Control bar. That is, among the two pairs, does one sample more closely match 227? – Group 1: 227 vs. 648

– Group 2: 227 vs. 950


April 2013

Grain Flavor


Grain Flavor


Training Conclusions

• Focus on: – Panelists indicating first attribute of dominance

immediately upon perception

– Chewing/swallowing procedure and evaluating dominance at the same time

– Clarity between attributes – nutty vs. grainy and sweet vs. caramelized especially as they may overlap

– Agreement among panelists regarding changes in dominant attribute – how many is expected


0 10 20 30 40

0.0

0.2

0.4

0.6

0.8

1.0

TDS Panel Average (smoothed data) - Product 227

Time (seconds)

Do

min

an

ce

(ra

te)

chance

significance

Caramelized Flavor

Dried Fruit Flavor

Grain Flavor

Nutty Flavor

Sweet

TDS – Temporal Dominance


0 10 20 30 40

0.0

0.2

0.4

0.6

0.8

1.0


Time (seconds)

Do

min

an

ce

(ra

te)

chance

significance

Caramelized Flavor

Dried Fruit Flavor

Grain Flavor

Nutty Flavor

Sweet



TDS Blue Group Results

• During 20 second chew: – 227 showed dominance progression of Grain Nutty

Caramelized

– 950 not clear for Grain initially, co-dominant with Sweet then Caramelized Nutty

• After swallow at 20 sec, 950 dominant for dried fruit and might consider lack of training or effect of heterogeneous product for 227 not showing dominance for Dried Fruit versus Sweet

• 227 showed Sweet and Grain linger, 950 showed Caramelized and Grain linger


0 10 20 30 40

0.0

0.2

0.4

0.6

0.8

1.0


Time (seconds)

Do

min

an

ce

(ra

te)

chance

significance

Caramelized Flavor

Dried Fruit Flavor

Grain Flavor

Nutty Flavor

Sweet



0 10 20 30 40

0.0

0.2

0.4

0.6

0.8

1.0


Time (seconds)

Do

min

an

ce

(ra

te)

chance

significance

Caramelized Flavor

Dried Fruit Flavor

Grain Flavor

Nutty Flavor

Sweet



TDS Red Group Results

• Chewing occurred for 20 Seconds, during that time:

– Both samples initially showed significance for Grain followed by Nutty

– 648 dominance for Nutty closer to the swallow at 20 sec.

• After swallow, both samples dominant for dried fruit

– Then 648 dominant changed quickly in dominance GrainSweetNutty however 227 dominant for GrainNutty

• Both samples lingered with significant Grain



• Both 648 and 950 show differences in TDS dominance profile versus 227

• Due to lack of training and replicate evaluations, data might not be stable enough to draw firm conclusions



Session Leaders: Suzanne Pecore General Mills Inc Peter Love Compusense Inc

Objectives of TOS Study

• To understand if the temporal profile has been affected by the reformulation of the E18 bar, and

• To identify a formulation option that more closely matches the Control bar. That is, among the two pairs, does one sample more closely match 227? – Group 1: 227 vs. 648

– Group 2: 227 vs. 950

Proportion Appearing First


• This plot helps to quickly see if differences were observed in the up-front attribute.

• Further data analyses can be focused among those attributes appearing above chance.

Proportion Appearing First Group 1: 227 vs. 648

• Grain flavor appears first across the eating experience in both bars.

• 227 shows some Sweet and Nutty appearing first for a few panelists.

• 648 shows Nutty appearing first in the 2nd bite, and Dried Fruit in aftertaste for some panelists.

Note: Data from untrained panelists attending seminar.

Proportion Appearing First Group 2: 227 vs. 950

• 950 shows more Caramelized in first two bites, and a strong hit of Grain flavor by the 3rd bite. 227 shows some Sweet in 2nd bite.

• Nutty is other attribute appearing above chance in both bars.


Weighted Order of Occurrence


• This plot shows calculated score for each bite, with higher rating indicating earlier appearance for that attribute. – For each panelist and each bite, the attribute appearing 1st is assigned

a ‘3’ in value, the attribute appearing 2nd = a ‘2’, and the attribute appearing 3rd = a ‘1’.

– For each attribute at each bite, these values are averaged to create the Weighted Order of Occurrence score.

• This weighted score can also be subjected to standard statistical analyses used with descriptive panel data, such as analysis of variance.

Weighted Order of Occurrence Group 1: 227 vs. 648

• Grain and Nutty appear early in both products across bites.

• Key differences are appearance of Sweet in 227 bar (significant sample effect in ANOVA), and Dried Fruit in 648 at 3rd Bite.


Sweet TOS Plot Group 1: 227 vs. 648


• Plot illustrates the earlier appearance of Sweet Taste in 227 across all three bites and greater intensity in the aftertaste.

Weighted Order of Occurrence Group 2: 227 vs. 950


• Similar to other pair, Grain and Nutty appear early in both products.

• Key difference is appearance of Caramelized in 950 bar (significant sample effect in ANOVA).

Caramelized TOS Plot Group 2: 227 vs. 950

• Plot illustrates the earlier appearance of Caramelized in 950 across all three bites and greater intensity in the aftertaste.




• Both options, 648 and 950, show TOS differences from sample 227.

• Due to lack of training among these panelists, data might not be stable enough to draw firm conclusions

Project Objective • The E18 Snack Bar is being reformulated to

improve its health credentials.

• Goal is to match the sensory characteristics of the current E18 Bar as closely as possible.

• How well did the each of the methods answer the question?

• Please draw your own conclusions.


Observations

• All temporal methods require training and practice to deliver optimal results.

• Each method provided unique temporal information that demonstrated difference amongst the products.

• The data collected in this seminar represent examples for illustrative purposes and do not represent data obtained from well-trained and experienced panelists.


April 2013

Flip Charts for the Breakout Rooms

Please note, these are transposed verbatim and would probably benefit from discussion and clear

attribution. I cannot be certain which method is being referred to in all cases.

1. Progressive Profiling

a. Needed spot for “Other”; plastic note in 227

b. A bit of delay in loading data. Love you can go back and change rating

c. During one Eval the screen froze, so invalid data (since while waiting expectorated etc.)

d. Yes. Where is “Other” – cardboard

e. Slow screen transition time could lead to faulty conclusions

f. Due to quick dissolution, technique may not fit to product form.

g. Bars very similar in taste & texture, so differentiation not apparent during the test

h. Memory may affect the rating

i. The accumulating may lead to a different 1st and 2nd bite.

j. Be very careful if want to use with untrained/consumers. Need to learn protocol.

k. Hard to have part left to expectorate because eval time longer.

l. Time transitions were too long – product disintegrated

m. Clearer instructions

n. Found additional flavors- Not able to express fully descriptive experience.

2. T O S

a. Expectorate/swallow need discussion (per objective)

b. No rinsing between smp (unused to)

c. Influenced by visual cues (nut=nutty;sticky=sweet)

d. Pre-portioning helpful/important

e. Lag time from perception to recording

f. Frustrated – not able to show intensity diffs (even tho doing D.A. separately)

g. 2 Aftertastes – couldn’t show

h. Aftertaste timing by computer good.

3. TDS

a. Can’t capture 2 simultaneously

b. Time between clips good to capture

c. Reps and panelists with such variable products

d. Profiling and clock captures it all

e. Rapid – no time to think

f. Need to consume more in TOS

g. Can see bite-to-bite diffs

h. Other is not consistent among panelists thus, NSD

i. Aftertaste dependent on the 4th bite

j. Does it work with very dominant flavor?

Temporal Methods Seminar - ASTM International · Presentation of the problem and logistics ... Aromatic Profile of Prototype E18 Bars ... reference is sucrose in water. Temporal Methods

Documents